| Summary: | 碩士 === 國立中山大學 === 資訊工程學系研究所 === 107 === Ad hoc is a main technique for non-infrastructure networks. The performance in ad hoc is affected by the design of routing algorithm. Earlier works consider the scenario that consists at least one end-to-end path between all pair-nodes, called MANETs (Mobile Ad hoc Networks). However, the nodes in network exist in the characteristics of clustering and periodicity, resulting numerous communities locating in network. The end-to-end path between each community is scarce. The traditional routing algorithm in MANETs fail in such scenario.
DTNs (Delay Tolerant Networks) utilize store-and-forward mechanism to transmit data. The nodes in DTN store packets into its buffer and waits for suitable relay. However, the buffer size is limited in realistic scenario. Selecting appropriate relay nodes and dropping messages are two keys to increase the buffer efficiency. In selecting relay nodes, the contact history can provide some useful information. However, most of previous works consider only single factor to evaluate the utility of a node. For example, the contact number between two nodes, periodic information, or community information. In realistic scenario, the nodes in a network usually enclose multiple characteristics. In this work, we consider the short time factor, long time factor, and the spreading ability of a node to evaluate the utility of a node, giving a comprehensive consideration of a node. In dropping policy, we adopt the similar concept as in selecting relay nodes to keep the higher utility packets in buffer.
In simulation, we adopt TVCM (Time Variant Community Mobility) as mobility model. The nodes in TVCM perform both periodicity and clustering. Furthermore, the movement data are demonstrated to be highly similar to realistic mining data. The results show that our proposed method performs better than previous works.
Keywords : Ad hoc networks, DTN ( Delay Tolerant Network), Store-and-Forward, Community, TVCM, Periodicity.
|
|---|
